Evaluation of Serum Des-Gamma-Carboxy Prothrombin for the Diagnosis of Hepatitis B Virus-Related Hepatocellular Carcinoma: A Meta-Analysis

Aim To explore the diagnostic efficacy of des-gamma-carboxy prothrombin (DCP) in hepatitis B virus- (HBV-) related hepatocellular carcinoma (HCC). Methods A retrospective study of 459 cases from June 2016 to March 2018 was undertaken, and records of the DCP levels were extracted. The sensitivity, specificity, and cutoff points were calculated using SPSS 17.0 software. A systematic search in PubMed, Web of Science, and the Cochrane Central Register of Controlled Trials was performed for articles published in English from 1997 to 2017, focusing on serum DCP for HBV-related HCC. Data on sensitivity, specificity, the positive likelihood ratio (PLR), negative likelihood ratio (NLR), and diagnostic odds ratio (DOR) were extracted from five studies by systematic search and one study of our own. The summary receiver operating characteristic (sROC) curve was obtained, and the area under the receiver operating characteristic (AUROC) curve was calculated. Results The pooled sensitivity, specificity, PLR, NLR, and DOR were 0.71 (95% CI: 0.59, 0.80), 0.93 (95% CI: 0.87, 0.96), 9.5 (95% CI: 5.2, 17.5), 0.32 (95% CI: 0.22, 0.46), and 30 (95% CI: 13, 72), respectively. The AUROC curve was 0.91 (95% CI: 0.88, 0.93). Conclusions In the diagnosis of HBV-related hepatocellular carcinoma (HCC), DCP is an ideal marker that should be considered for surveillance purposes.


Introduction
Primary liver cancer is the fifth most common cancer and is responsible for the second most common cancer-related deaths worldwide [1]. Hepatocellular carcinoma (HCC) alone accounts for approximately 80% of all cases [2] and is one type of malignancy with a very poor prognosis [1]. Early identification can improve the prognosis. However, HCC usually arises against the background of liver damage, and the tumors are usually relatively large before any symptoms become evident [3]. Previous studies have shown that if patients are diagnosed at an early stage, the 5-year survival rate can be above 70% [4]. If patients are diagnosed at a late stage, however, the 5-year survival rate is less than 5% [5]. These features make early identification both difficult and important. Ultrasonography (US) alone, without concurrent detection of serum alpha fetoprotein levels, has been recommended for the surveillance of HCC, according to the representative guidelines of the American Association for the Study of Liver Diseases in 2010 [6] and the European Association for the Study of the Liver in 2012 [7]. However, interpretation of the sonogram is dependent on the expertise of the operator and quality of the equipment [8] and can be disrupted by other conditions of the liver, such as cirrhosis [9]. The overall sensitivity of US in this context is only 0.593 [10].
The identification of new markers for the diagnosis of HCC is urgently required. Serum des-gamma-carboxy prothrombin (DCP), also known as prothrombin-induced by vitamin K absence-II (PIVKA-II), was first reported by Liebman et al. in 1984 [11]. Its value has been confirmed in the diagnosis of HCC in a series of clinical trials [12][13][14][15][16][17][18][19][20]. However, the sensitivity, specificity, and cutoff points in previous studies have been inconsistent, and in some instances, even conflicting [21]. One of the reasons for this disparity could be the differences in etiology and the fact that diagnostic values were mainly derived from various patients without homogeneous etiologies.
Hepatitis B virus (HBV) infection is the main causative factor of HCC, and more than half of all cases could be attributed to HBV infection worldwide [22]. Even with improvements in technologies, the morbidity and mortality of HBV-related HCC have still shown a steady increase [23].
However, the precise diagnostic efficacy has never been fully evaluated. We conducted this meta-analysis to elucidate this pertinent issue.  Five studies were mice-related, and thus excluded.

Methods
Three studies were proceedingsreports, and were thus excluded.
31 studies were focused on other types of HCC (12 studies of HCV-related HCC), and were thus excluded.
10 studies were related to prediction of the condition instead of its diagnosis, and were thus excluded.
60 studies did not evaluate the utility of DCP, and were thus excluded.
Six studies included for meta-analysis.
18 studies were duplications, and thus excluded.
Six studies were reviews, and thus excluded.
Five studies selected from the search.
One study of our own included.
144 studies identifiedthrough database search (23from PubMed and 121 from Web of Science

) (3) AND (2) AND (1): limited in the English language
Altogether, 137 studies were included. Two studies were excluded because they were not written in English, one was excluded because of duplication, four more were excluded because they reported on the proceedings of various conferences, and nine were excluded because they were reviews. Thus, 121 articles were included after the first level screening. (3) (DCP) or (descarboxyprothrombin) or (des gamma-carboxy prothrombin) or (des-gammacarboxy prothrombin) or (des-gamma-carboxyprothrombin) or (decarboxyprothrombin) or (non-carboxylated factor II) or (PIVKA II) or (PIVKA-II) or (protein induced by vitamin K absence or antagonist-II) or (acarboxyl prothrombin) or (decarboxylated prothrombin) or (protein induced by vitamin K absence or antagonists) or (PIVKA) or (prothrombin precursor) or (Isoprothrombin) or (des γ-carboxy prothrombin) or (des-γ-carboxy prothrombin) or (des-γ-carboxyprothrombin) (1) and (2) and (3) Altogether, 50 studies were included. 42 studies were reviews, and eight studies were unrelated, after screening of the title and abstract. Thus, no articles were included after the first-level screening.
(3) Level 3. Altogether, 126 studies were included and six studies were excluded because they comprised reviews.
(4) Level 4. Five studies were eventually included from the searches conducted. Six studies were excluded because they were case reports; five studies were excluded because they were mice-related; three studies were excluded because they reported the proceedings of various conferences; 31 studies were excluded because they reported on other types of HCC (12 studies of hepatitis C virus-(HCV-) related HCC); 10 studies were excluded because they were related to the prediction of the condition, instead of its diagnosis; and 60 studies were excluded because they did not evaluate the utility of DCP.

Data Extraction.
We extracted data on the authors, year of publication, country, number of patients, sensitivity, specificity, cutoff points, and area under the receiver operating characteristic (AUROC) curve from the selected studies (Table 2).

Assessment of Methodological
Quality. The quality of each study was assessed according to the QUADAS (quality assessment of studies of diagnostic accuracy included in systematic reviews) checklist. Each of the 14 items in the QUADAS checklists was as "yes," "no," or "unclear" [30]. Two items were not assessed, one of which was "Were uninterpretable/intermediate results reported?" This question was not assessed because the concentration of DCP was determined by test kits; thus, there were no uninterpretable/intermediate results. The other unassessed item was that of "Withdrawals explained?" This item was not assessed because all the retrospective studies were evaluated without consideration of withdrawals. Twelve items of the QUADAS checklist are shown, and the quality of the studies are presented in Table 3.

Data Analysis.
For the retrospective evaluation of our previous study, the analysis of the receiver operating characteristic (ROC) curve was applied to measure diagnostic efficacy. The cutoff points and AUROC were calculated using SPSS 17.0 software. For the meta-analysis, a funnel plot was constructed and P values were calculated. Publication bias existed when a P < 0 05 was observed. The pooled sensitivity, specificity, PLR, NLR, and DOR were summarized, and the AUROC was calculated. The statistical analyses of the meta-analysis were accomplished, using Stata 13.1 software.

Characteristic of the Selected Studies.
A total of 144 studies were identified from the searches conducted, of which five [9,24,[31][32][33] studies were considered suitable for inclusion in the meta-analysis, after exclusion of reviews, case reports, duplications, conference proceedings, and other unsuitable studies. All included studies were published in English. Another suitable study of ours conducted on 459 patients was included. As shown in Table 1, 2472 patients from six studies were eventually included for the meta-analysis.

Quality of the Studies.
The results of the QUADAS assessment are presented in Table 2. All included studies adopted a retrospective design, and this was not regarded as the representative of the patient spectrum. In addition, three [24,31,32] studies did not clearly describe the exclusion criteria applied. However, other items were effectively reported, and all six studies scored an "A."

Analysis of Publication Bias.
A funnel plot was used to examine publication bias. As shown in Figure 2, the P value was 0.92, indicating the absence of publication bias among the six studies.

Heterogeneity
Analysis. The analysis of heterogeneity is presented in Figure 3. The studies could have been incorporated with an index of correlation (mixed model) of 0.33, and the proportion of heterogeneity was likely due to the threshold effect, which was 0.11.

Discussion
The meta-analysis of the six studies confirmed the diagnostic efficacy of DCP in patients with HBV-related HCC. All six studies were published recently (three being published in 2017). In addition, each of the studies included in the metaanalysis was of high quality and scored an "A" according to the QUADAS checklist. All serum samples were measured by enzyme immunoassay; however, limitations must still be  acknowledged. Firstly, all six studies were retrospective, which could increase the representative patient spectrum. Secondly, the cutoff points differed among the studies. The 40 mAU/mL cutoff point in the serum had been established in the population with heterogeneous etiologies [19]. The same cutoff point of 40 mAU/mL was employed in three studies [9,24,31] (two from Korea and one from China), whereas the cutoff points of 32.09 mAU/mL [32] and 40.5 m AU/mL [33] were employed in the other two studies (both from China). The study with the lowest cutoff point of 32.09 mAU/mL [32] investigated early-stage HBV-related HCC. In our own study, the cutoff point was 49.05 mAU/ mL. This value was the highest among the six studies. One of the main reasons for this relatively high cutoff point was the fact that all HCC cases in our study were diagnosed by histological examination. In comparison with other studies, in which diagnosis was made by histological examination or imaging characteristics, histological examination was considered the gold standard with the highest diagnostic sensitivity and specificity for all cancers' diagnosis [34]. The imaging diagnosis also takes an important role in diagnosis, but sometimes it is discordant with histological examination [35], especially in early HCC [36]. As a general rule, due to the high negative predictive value, reference values defining "below the cutoff point" biomarker concentrations (such as the 99th percentile of a healthy reference population with a coefficient of variation of <10%) are especially useful for diagnostic purposes [37]. With the gold standard of histological examination, the highest cutoff point of our study represents the most accurate level of DCP in HBV-related HCC. DCP is produced by HCC and can conversely stimulate the growth and invasion of HCC through different signal pathways [38]. The diagnostic efficacy of histological examination has been confirmed for HCC. However, most of the previous studies were conducted in populations with heterogeneous liver diseases, and the respective sensitivities, specificities, and cutoff points were inconsistent among the various studies [21]. The 40 mAU/mL cutoff point has also been confirmed in populations with heterogeneous etiologies predominantly infected with HCV [12][13][14][15][16][17][18][19][20]39]. The present study, to our knowledge, is the first to report a meta-analysis that evaluates the diagnostic efficacy of DCP in the detection of HCC with homogeneous etiologies (HBV infection).
The sensitivity, specificity, PLR, NLR, DOR, and AUROC are all indices of diagnostic efficacy. The values of sensitivity, specificity, and the AUROC that are closer to 1 are preferred. A likelihood ratio greater than 1 indicates that the test result is associated with the disease. The value of a DOR ranges from 0 to infinity, and the higher values indicate better diagnostic efficacy. In the present meta-analysis, we included six studies of 2472 HBV-infected patients from China and Korea. We found that in the HBV-infected population, the pooled sensitivity, specificity, PLR, NLR, and DOR were 0.71, 0.93, 9.5, 0.32, and 30, respectively. The AUROC was as high as 0.91. The AUROC of DCP were 0.89, 0.797, and 0.893 in three meta-analyses of predominantly HCVinfected populations [29,40,41]. In comparison with the meta-analysis of populations with heterogeneous etiologies, the AUROC value was higher in HBV-related HCC. The diagnostic efficacy of DCP was more favorable in HBVrelated HCC.
In various guidelines for HCC worldwide, DCP has only been recommended for the surveillance of HCC. For instance, in the 2013 guidelines of Japan [8], it is recommended for the surveillance of HCC caused mainly by HCV (68%) [42]. The present results show the superior diagnostic efficacy of DCP in HBV-related HCC. In view of these findings, we suggest that DCP should be considered for the surveillance of HCC in the established guidelines of other countries and regions, especially those with a high incidence of HBV infections, such as East Asia (except Japan) and Africa [43].
Ultrasonography has been recommended for the surveillance of HCC in almost all established guidelines worldwide [8]. However, the overall sensitivity of US is only 0.593 [10]. With a sensitivity of 0.71 and specificity of 0.93, DCP seems to be a more favorable choice than US.
With its high diagnostic efficacy, ease of use, and reproducibility, as well as its objectivity and noninvasiveness, DCP is an ideal marker that could be considered an easily accessible complement of US for the surveillance for HBVrelated HCC. As this meta-analysis was based on six retrospective studies, more prospective studies with relatively large samples are needed in the future.